Welcome to the Chair of Systems Design
At the Chair of Systems Design, ETH Zurich, we perform interdisciplinary Complex Systems Research with a particular focus on the understanding and modeling of phenomena present in social, socio-technical and socio-economic systems. To do so we apply and develop quantitative methods from statistical physics, applied mathematics, computer science and beyond.
In our new preprint, we use a large dataset about firm alliances, and we identify two distinct behaviors in the alliance formation process. By following the ``career path'' of all individual firms we find that newcomers, and nodes of low centrality, initially form links to nodes of similar or higher centrality, whereas established nodes switch their strategy and preferably form links to less central nodes. This ``universal'' behavior could serve as a strategy to avoid competition with peers and to access peripheral resources. Providing a network growth model based on k-core centrality, we show that the behavioral shift results from the abundance of choice among newcomer nodes.
By combining multiple social media datasets, it is possible to gain insight into each dataset that goes beyond what could be obtained with either individually. In this paper we combine user-centric data from Twitter with video-centric data from YouTube to build a rich picture of who watches and shares what on YouTube. We study 87K Twitter users, 5.6 million YouTube videos and 15 million video sharing events from user-, video- and sharing-event-centric perspectives. We show that features of Twitter users correlate with YouTube features and sharing-related features. For example, urban users are quicker to share than rural users. We find a superlinear relationship between initial Twitter shares and the final amounts of views. We discover that Twitter activity metrics play more role in video popularity than mere amount of followers. We also reveal the existence of correlated behavior concerning the time between video creation and sharing within certain timescales, showing the time onset for a coherent response, and the time limit after which collective responses are extremely unlikely. Response times depend on the category of the video, suggesting Twitter video sharing is highly dependent on the video content. To the best of our knowledge, this is the first large-scale study combining YouTube and Twitter data, and it reveals novel, detailed insights into who watches (and shares) what on YouTube, and when.
You do good scientific work but nobody cites you? Maybe this is because you are not well enough embedded in the scientific social network. In our new paper Predicting Scientific Success Based on Coauthorship Networks (arXiv:1402.7268) we take a look at the question what role mechanisms of social influence play for scientific success. Precisely, we quantify to what extent metrics of centrality of authors in the coauthorship network correlate with citation numbers of their papers. Surprisingly, we find that a machine learning classifier is able to predict whether a paper will be successful or not with a precision of 60 % (and a recall of 18%), based solely on metrics of social network centrality of its authors. This result clearly challenges the perception of "citations" as objective measures of scientific quality.
The recent crisis has brought to the fore a crucial question that remains still open: what would be the optimal architecture of financial systems? We investigate the stability of several benchmark topologies in a simple default cascading dynamics in bank networks. We analyze the interplay of several crucial drivers, i.e., network topology, banks' capital ratios, market illiquidity, and random vs targeted shocks. We find that, in general, topology matters only – but substantially – when the market is illiquid. No single topology is always superior to others. In particular, scale-free networks can be both more robust and more fragile than homogeneous architectures. This finding has important policy implications. We also apply our methodology to a comprehensive dataset of an interbank market from 1999 to 2011.
The latest results of a 5-year collaboration between our group and conservation biologists at the University of Greifswald have been published in Naturwissenschaften. In this new study, we show that a wild colony of bats can flexibly adapt its social structure in response to a dramatic population decline.
Our article titled "Gender Asymmetries in Reality and Fiction: The Bechdel Test of Social Media" has been accepted as a full paper in the 8th International AAAI Conference on Weblogs and Social Media (ICWSM '14).
The book "Collective Emotions: Perspectives from Psychology, Philosophy, and Sociology" edited by Christian von Scheve, Mikko Salmella is published by Oxford University Press. Chapter 26 "Modeling collective emotions in online social systems" by D.Garcia, A.Garas, and F.Schweitzer gives a nice overview of our research in this area.
Our recent work on the automatic re-modularization of software has been accepted to the International Conference on Modularity 2014. We use a stochastic strategy based on move refactoring and show that the worse the modularity of a given architecture, the better the improvement achieved by our approach.
Together with evolutionary biologists at the University of Zurich, we have published a new manuscript on an information-theoretic approach to coupling and leadership in animal groups. In this arXiv preprint, we illustrate the use of this technique on groups of wild meerkats carrying GPS collars.
We proudly announce that our recent work "Betweenness preference: Quantifying correlations in the topological dynamics of temporal networks" has been accepted for publication in the journal Physical Review Letters.
Going beyond the mere aggregate topology or activities of nodes in dynamic networks, in our paper we uncovered a so far unknown temporal-topological dimension of network dynamics. We further show that this new dimension crucially influences dynamical processes like for instance information diffusion or the spreading of diseases.
For a number of complex systems, a simple abstraction of their organization in terms of networks is not sufficient for understanding their structure, dynamics, and function. This observation raises fundamental questions: When are simple network models sufficient and when are they not? What additional ingredients are needed to accurately model the dynamical processes? With access to more and more relational data, what are the most efficient ways to capture the structural information?
These are questions that we would like to address in aour workshop on Higher-Order Models in Network Science, which is co-organized by Dr. Renaud Lambiotte, Dr. Martin Rosvall and Dr. Ingo Scholtes. The satellite will be co-located with NetSci 2014, the premiere international conference on complex networks. It be held on Tuesday, June 3rd, 2014 at University of California, Berkeley.
Today, Dr. Ingo Scholtes will give a talk on our recent work Categorizing Bugs with Social Networks at SE2014, the most important software engineering conference held in the german-speaking countries. Contributions to SE 2014 were on an invitation-only basis and required a previous publication at one of the international top-notch software engineering conferences. We are proud that the organizers identified our work as one of the last years' top contributions to the field of software engineering.
In our work, we quantitatively studied the importance of social structures on distributed software engineering processes. In particular, we combine Big Data techniques, network analysis and predictive analytics to automatically assess bug report quality. Our method can be used to improve the bug handling processes of large-scale Open Source Software communities. This work is an outcome of our research line of social software engineering and has been funded by the SNF in the context of a research project on distributed software engineering.